Researchers from the Department of Urology at Fudan University Shanghai Cancer Center evaluated artificial intelligence (AI)–assisted communications in the preoperative setting to assess its impact on patient anxiety and clinician workload. The performance of the model was assessed in a prospective, randomized, single-blinded phase II trial including patients with prostate cancer. Results were presented at the 2026 ASCO Genitourinary Cancers Symposium.1
“AI-assisted preoperative communication significantly reduces patient anxiety, enhances disease perception, and decreases clinician workload, identifying previously unaddressed patient concerns,” lead author Zheng Liu, MD, PhD, and co-authors wrote in their abstract.
The study enrolled patients with newly diagnosed prostate cancer who were scheduled for a radical prostatectomy. Patients were randomly assigned by room level to receive either standard communications (n = 130) or AI-assisted communications (n = 138) prior to their surgery.
Both groups received baseline evaluations for anxiety, emotion, and disease recognition, including GAD-7, VAS-A, I-PANAS-SF, and B-IPQ questionnaires. Specific questions from the assessments were collected by the AI system.
Personalized responses were generated by the large language model–based AI system for patients in the AI-assisted group, and both groups took the evaluations again after receiving their separate communications. Physician blinding was maintained during face-to-face interactions to account for potential confounding factors. The doctors also completed their own evaluation, NASA-TLX, to assess the workload impact of these communications.
The study measured changes in the evaluation scores between the two groups, focused especially on areas of anxiety with the GAD-7 questionnaire and physician workload with the NASA-TLX questionnaire. In addition to looking at changes in scores for the other assessments, the researchers also looked at differences in the preoperative communication time between the two groups.
A total of 3,507 questions were collected from patients between April and November 2025 for the study, or 13.08 questions per patient.
Patients in the AI-assisted communications group reported significantly lower anxiety than those in the control group, with a median GAD-7 score of 3.5 (interquartile range [IQR] = 2.0–5.5) vs 6.5 (IQR = 4.5–8.5; P < .001).
Reduced physician workload as well as shorter communication time was also observed with the large language model–based patient communication system. Physicians reported a NASA-TLX score of 37.0 for the AI-assisted communications group (IQR = 27.0–46.0) vs 54.0 (IQR = 35.0–73.0) for the control group (P < .001). Communications took an average of 10.8 minutes for the AI-assisted group as compared with 19.4 months for the control group (P < .001).
Additionally, the use of AI-assisted communications led to more favorable emotional profiles, lower negative illness recognition, and higher satisfaction overall.
An exploratory analysis showed that AI assistance alone without physician consultation could even improve patients’ anxiety and emotional outcomes.
Going forward, the researchers intend to identify queries that are not appropriate for the AI system to answer and redirect them to physicians, thereby further reducing manual review and workload burden. Additionally, they will explore the generalizability of the AI system across languages, cultures, and health-care settings.
DISCLOSURE: The study authors had no related relationships to disclose.
REFERENCE
1. Liu Z, Xu H, Hong Z, et al: Assistance of a large language model–based AI agent in preoperative communication for reduction of prostate cancer patients’ anxiety and clinicians’ burden: A prospective, randomized, single-blinded, phase II trial. 2026 ASCO Genitourinary Cancers Symposium. Abstract 56. Presented February 27, 2026.
